On Mon, 04 Apr 2011 16:33:05 +0200 Luk Claes <luk@xxxxxxxxxx> wrote: > Hi > > A Debian user reported a bug that a slash gets appended to the > filesystem when he mounts a share. This only happens when mtab is > symlinked to proc which I can reproduce: > > $ grep cifs /etc/fstab > //<host>/<path> /mnt cifs user,user=<username> 0 0 > > $ mount /mnt > > $ grep cifs /etc/mtab > //<host>/<path>/ /mnt cifs > rw,nosuid,nodev,relatime,unc=\\<host>\<path>,username=<username>,uid=1000,forceuid,gid=1000,forcegid,addr=<ip>,file_mode=0755,dir_mode=0755,serverino,rsize=16384,wsize=57344 > 0 0 > > $ umount /mnt > umount: /mnt mount disagrees with the fstab > > There does not seem to be any other filesystems that append a slash to > the filesystem in proc. Does cifsfs do that intentionally (in which case > the client should cope with it: by means of umount.cifs?) or should it > get fixed in the kernel (maybe the client should still cope with it)? > > Can you give me a hand in identifying the code that would need to be > updated to fix this so I can prepare a patch? > > Cheers > > Luk mount.cifs will convert the delimiters in the "device string" to forward slashes (in case someone specifies a UNC with backslashes) and then passes it down to the kernel. When it does this, it adds an unnecessary '/' to the end. Probably what we need to do is make it pass down the "original" device string to the kernel mount() call. The reason mount.cifs does this though is because /proc/mounts turns '\\' into an escaped number string. So, in addition to changing mount.cifs to fix this, the kernel routines that format /proc/mounts should also be fixed to handle backslashes correctly. Cheers, -- Jeff Layton <jlayton@xxxxxxxxxx> -- To unsubscribe from this list: send the line "unsubscribe linux-cifs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
